Method

Three layers, in order. Each is built on the one before it, and each is progressively less certain — the first is measurement, the second is diagnosis, the third is conditional projection.

IMF World Economic Outlook — via the DataMapper API

The macro backbone. Covers 172 economies from 1980 to 2031, and critically it includes the IMF’s own published forecasts to 2031 — so the baseline the scenarios perturb is an institutional forecast, not something this model invented.

• GDP, current prices NGDPD · US$ bn
• Real GDP growth NGDP_RPCH · %
• GDP per capita, current prices NGDPDPC · US$
• GDP, PPP PPPGDP · Intl$ bn
• Share of world GDP (PPP) PPPSH · %
• Inflation, average consumer prices PCPIPCH · %
• General government gross debt GGXWDG_NGDP · % of GDP
• General government net lending/borrowing GGXCNL_NGDP · % of GDP
• Current account balance BCA_NGDPD · % of GDP
• Population LP · millions

World Bank World Development Indicators

The balance-sheet stocks. These are annual and lag the IMF data by a year or two for some economies, which is why the model always reports the year each reading comes from rather than assuming it is current.

• Listed equity market capitalisation CM.MKT.LCAP.GD.ZS · % of GDP
• Domestic credit to private sector FS.AST.PRVT.GD.ZS · % of GDP
• Broad money FM.LBL.BMNY.GD.ZS · % of GDP
• Gross capital formation NE.GDI.TOTL.ZS · % of GDP
• Gross national savings NY.GNS.ICTR.ZS · % of GDP

Derived

• Real GDP, constant 2025 US$ — Derived: IMF NGDPD 2025 chained by NGDP_RPCH

Real GDP is built by anchoring on each economy’s 2025 nominal dollar GDP and chaining backward and forward through the IMF’s real growth rates. This is what makes adding economies into continents legitimate: a sum of constant-dollar series is a real aggregate, whereas a sum of nominal series confuses growth with inflation and currency moves.

Four stocks, each as a percentage of GDP. For each one the model takes the economy’s own mean and standard deviation over 2000–2024, and expresses the latest reading as a z-score — how many standard deviations above or below its own normal that economy is now carrying. The composite is the weighted mean of those z-scores.

Bands. Above +1.5σ is called stretched, +0.5 to +1.5 elevated, −0.5 to +0.5 balanced, below −0.5 subdued. These cut points are a presentational choice, not a finding — the underlying number is continuous and is always shown next to the label.

Each pathway is a set of deltas applied to the IMF’s own baseline, phased in over two or three years. Growth, inflation, the real interest rate and an asset-valuation drift are set per pathway; everything else is consequence. Government debt in particular is not assumed — it falls out of the standard debt-dynamics identity:

Δ(D/Y) = (r − g)/(1 + g) · (D/Y) − primary balance

  D/Y  government debt as a share of GDP
  r    nominal effective interest rate = scenario real rate + inflation
  g    nominal GDP growth = (1 + real growth)(1 + inflation) − 1

When nominal growth exceeds the interest rate, an economy can run deficits without the debt ratio rising. When it does not, the ratio climbs unless the primary balance improves. That relationship is the single most useful thing in the model, and it is arithmetic rather than opinion.

Growth and inflation are deltas against the IMF baseline. The real rate is a level, not a delta, and is the weakest assumption in the model — market interest rates are not in the free data feeds, so the model sets them by pathway rather than observing them.

• It is not the MGI balance sheet. MGI builds national balance sheets from national accounts — including unlisted equity, household real estate, land, and the full financial-sector layer. That underlying dataset is not public. This model follows the same method — stocks as GDP multiples, judged against their own history — on the public series that are available. The levels therefore differ: MGI puts US equity at 3.7× GDP including unlisted firms, while listed market capitalisation in the World Bank series is about 2.2×. Compare directions and gaps here, not levels against MGI’s.
• There is no household wealth series. The wealth index in the scenarios is a proxy — real growth plus the pathway’s valuation drift — not measured household net worth. It shows the direction and rough magnitude of real wealth under each pathway, nothing finer.
• Interest rates are assumed. Each pathway carries a real rate rather than an observed yield curve. Since the debt identity is sensitive to r − g, debt paths are sensitive to this choice.
• Aggregation hides distribution. A continent’s growth rate says nothing about the spread inside it, and national wealth says nothing about who holds it. MGI’s own note is worth repeating: US average per-capita wealth is dominated by the top of the distribution, and the bottom half holds less per person than the average household in most other rich countries.
• Scenario probabilities are yours. The weights on the scenarios page are an input for you to set. The model has no view on which pathway is more likely and does not attempt one.
• Shocks are not modelled. Wars, pandemics, defaults, currency crises and political ruptures do not appear. The pathways are smooth; history is not.

Continents follow geography, with two conventions worth naming: Russia is grouped with Europe and Turkey with Asia. Blocs overlap by design — an economy can be in the G7, the OECD and the EU at once. The 23 economies in MGI’s own balance-sheet sample are available as a bloc so you can view the same group MGI reports on.

The diagnostic and the four pathways follow McKinsey Global Institute, The global balance sheet 2026: Imbalance and divergence (July 2026). That report supplies the idea that an economy has a balance sheet worth reading alongside its income statement, the test of whether wealth growth is backed by real capital formation or by asset prices, the four resolution pathways with their historical analogies, and the economy-specific swing factors that appear as levers on the scenarios page.

What is reproduced here is the framework, computed independently from open data. Where this model and MGI agree — the United States and China carrying stretched balance sheets, the major eurozone economies closer to balance, Japan elevated — that agreement is a genuine cross-check, because the inputs are not the same. Where they disagree, MGI has the better data and this model has the advantage of being refreshable by anyone in a single command.